Ranking Distillation Suite (RD-Suite)

RD-Suite数据集由谷歌山景城的研究团队创建，旨在为排序蒸馏模型提供一个系统化和统一的基准。该数据集包含四个大型真实世界数据集，分别涉及文本和数值两种主要模态，以及标准蒸馏和蒸馏转移两种应用场景。数据集的创建过程强调了通用性、简单性和可重复性，以确保研究的可比性和有效性。RD-Suite数据集的应用领域主要集中在搜索引擎和推荐系统等用户界面相关的技术中，旨在通过蒸馏技术提高模型的性能和效率。

The RD-Suite dataset was created by the research team at Google Mountain View, aiming to provide a systematic and unified benchmark for ranking distillation models. This dataset includes four large real-world datasets, covering two primary modalities: text and numerical data, as well as two application scenarios: standard distillation and distillation transfer. The construction of the RD-Suite dataset emphasizes generality, simplicity, and reproducibility to ensure the comparability and validity of related research. The RD-Suite dataset is mainly applied in user interface-related technologies such as search engines and recommendation systems, with the goal of improving model performance and efficiency through distillation techniques.